This face following robot project realises real time following of faces and human bodies by the robot in different environments through the integrated application of HOG (Histogram of Oriented Gradients) algorithms, video processing, servo control and sensor technologies. The project first establishes a clear workflow, including the steps of video head startup, image processing, servo control, and distance sensing. Through different and multiple experiments, the parameters of the HOG algorithm were adjusted, including the tolerance value, cell size and block size, these data improved the accuracy of detecting human targets and also improved the accuracy of the robot's work. Although the project achieved satisfactory results, there is still have some improvement. It is recommended to optimise the algorithms to make real time detection efficiency , and also increase flexibility of the robot in different surroundings, and to consider the introduction of more advanced hardware and machine learning techniques to improve system performance. Overall, the project provides useful experience in applying the HOG algorithm to real face following robot projects, and also suggests potential directions for future improvements.